Genomic patterns of pleiotropy and the evolution of complexity Zhi Wanga,1, Ben-Yang Liaob, and Jianzhi Zhanga,2 aDepartment of Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, MI 48109; and bDivision of Biostatistics and Bioinformatics, Institute of Population Health Sciences, National Health Research Institutes, Miaoli County 350, Taiwan, Republic of China Edited by Günter P. Wagner, Yale University, New Haven, CT, and accepted by the Editorial Board August 31, 2010 (received for review April 6, 2010) Pleiotropy refers to the phenomenon of a single mutation or gene hibit the origins of complex organisms and hence is puzzling to affecting multiple distinct phenotypic traits and has broad impli- evolutionary biologists (19, 20). cations in many areas of biology. Due to its central importance, Although pleiotropy has been examined in detail in a few genes pleiotropy has also been extensively modeled, albeit with virtually (21, 22), its genomic pattern is largely unknown, which seriously no empirical basis. Analyzing phenotypes of large numbers of limits us from evaluating mathematical models of pleiotropy, ver- yeast, nematode, and mouse mutants, we here describe the geno- ifying theoretical inferences from these models, and testing various mic patterns of pleiotropy. We show that the fraction of traits pleiotropy-related hypotheses in many fields of biology. In this altered appreciably by the deletion of a gene is minute for most work, we compile from existing literature and databases pheno- genes and the gene–trait relationship is highly modular. The stan- types of large numbers of yeast, nematode, and mouse mutants. We dardized size of the phenotypic effect of a gene on a trait is ap- describe the genomic patterns of pleiotropy in these organisms and proximately normally distributed with variable SDs for different show that these patterns drastically differ from any mathematical genes, which gives rise to the surprising observation of a larger model of pleiotropy. We further demonstrate that the cost of per-trait effect for genes affecting more traits. This scaling prop- complexity is substantially alleviated when the empirical patterns erty counteracts the pleiotropy-associated reduction in adaptation of pleiotropy are taken into consideration and that the observed “ ” rate (i.e., the cost of complexity ) in a nonlinear fashion, resulting value of a key parameter of pleiotropy falls in a narrow range that in the highest adaptation rate for organisms of intermediate com- maximizes the optimal complexity. EVOLUTION plexity rather than low complexity. Intriguingly, the observed scal- ing exponent falls in a narrow range that maximizes the optimal Results and Discussion complexity. Together, the genome-wide observations of overall Most Genes Affect only a Small Fraction of Traits. To uncover the low pleiotropy, high modularity, and larger per-trait effects from genomic patterns of pleiotropy, we compiled three large datasets genes of higher pleiotropy necessitate major revisions of theoret- of gene pleiotropy for the baker’s yeast Saccharomyces cerevisiae, ical models of pleiotropy and suggest that pleiotropy has not only one for the nematode worm Caenorhabditis elegans, and one for allowed but also promoted the evolution of complexity. the house mouse Mus musculus (SI Methods). The first dataset, yeast morphological pleiotropy, is based on the measures of 279 genetics | adaptation | modularity | yeast | phenotype morphological traits in haploid wild-type cells and 4,718 haploid mutant strains that each lack a different nonessential gene (23). leiotropy occurs when a single mutation or gene affects mul- The second dataset, yeast environmental pleiotropy, is based on Ptiple distinct phenotypic traits (1). Pleiotropy has broad impli- the growth rates of the same collection of yeast mutants relative to cations in genetics (1–3), development (4, 5), senescence (6), dis- the wild type in 22 different environments (24). The third dataset, ease (7, 8), and many evolutionary processes such as adaptation (2, yeast physiological pleiotropy, is based on 120 literature-curated 9–13), maintenance of sex (14), preservation of redundancy (15), physiological functions of genes recorded in the Comprehensive and stabilization of cooperation (16). For example, the antago- Yeast Genome Database (CYGD). The fourth dataset, nematode nistic pleiotropy theory of senescence asserts that alleles beneficial pleiotropy, is based on the phenotypes of 44 early embryogenesis to development and reproduction are deleterious after the re- traits in C. elegans treated with genome-wide RNA-mediated in- productive age and cause senescence, which may explain why all terference (25). The fifth dataset, mouse pleiotropy, is based on species have a limited life span (6). Pleiotropy is the main theo- the phenotypes of 308 morphological and physiological traits in retical reason behind the hypothesis that morphological evolution gene-knockout mice recorded in Mouse Genome Informatics occurs more frequently through cis-regulatory changes than (MGI). These five datasets provide qualitative information about through protein sequence changes (9), as the former are thought to the traits that are affected appreciably by each gene. In addition, be less pleiotropic than the latter. Pleiotropy also has important the first dataset also includes the quantitative information of the implications in human disease, because many genetic defects each effect size of each gene on each trait. Even after the removal of affect multiple phenotypic traits. For instance, mutations in the genes that do not affect any trait and traits that are not affected by homeobox gene ARX cause ambiguous genitalia and lissencephaly any gene, these five datasets each include hundreds to thousands of (whole or parts of the surface of the brain appear smooth) (Online genes and tens to hundreds of traits (Fig. 1). They are thus suitable Mendelian Inheritance in Man no. 300215). for examining genome-wide patterns of pleiotropy. Due to pleiotropy’s central importance in biology, several mathematical models of pleiotropy have been developed and im- portant theoretical results have been derived from the analyses of Author contributions: Z.W. and J.Z. designed research; Z.W. and B.-Y.L. performed re- these models (10, 11, 17, 18). For example, Fisher proposed that search; Z.W. and J.Z. analyzed data; and Z.W. and J.Z. wrote the paper. every mutation affects every trait and the effect size of a mutation The authors declare no conflict of interest. on a trait is uniformly distributed (10). On the basis of this model This article is a PNAS Direct Submission. G.P.W. is a guest editor invited by the Editorial and the assumption that the total effect size of a mutation is con- Board. stant in different organisms, Orr derived that the rate of adaptation 1Present address: Sage Bionetworks, Seattle, WA 98109. of a population to an environment quickly declines with the in- 2To whom correspondence should be addressed. E-mail: [email protected]. fi crease of the organismal complexity, which is de ned by the total This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. number of traits (12). This “cost of complexity” would likely pro- 1073/pnas.1004666107/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1004666107 PNAS Early Edition | 1of6 Downloaded by guest on September 27, 2021 In all five datasets, we observed that most genes affect only rewired networks, which have randomized links but an unchanged a small fraction of traits and only a minority of genes affect many number of links per node (32) (Fig. 2C). We then calculated the traits (Fig. 1). The median degree of pleiotropy varies from one to scaled modularity of a network, which is the difference between seven traits (or 1–9% of the traits considered) in these datasets. the observed modularity of a network and the mean modularity of The degree of pleiotropy measured by the percentage of examined its randomly rewired networks in terms of the number of SDs (32) traits is expected to be more accurate for the three datasets in (Methods). Our results show large scaled modularity (34–238) in which the same set of traits was examined in all mutants (Fig. 1 A, each of the five gene–trait networks examined (Fig. 2 D–H), pro- B,andD). By bootstrapping traits, we found that the SDs of our viding definitive evidence for the modular pleiotropy hypothesis. estimated median and mean degrees of pleiotropy are generally Our results remain qualitatively unchanged even when a random small, indicating that these estimates are precise (Fig. 1). To ex- 50% of the traits in each dataset are removed (Table S3). The amine the impact of the number of examined traits on the esti- modularity would be overestimated if the genetic correlations mated pleiotropy, we randomly removed 50 and 90% of the traits among traits are biased upward in our datasets compared with the from each dataset, respectively. We found that the mean and complete datasets that include all possible traits. Although we do median degrees of pleiotropy, measured by the percentage of traits not know whether this bias exists, to be conservative, we merged examined, remain largely unchanged (Table S1), suggesting that traits whose genetic correlation coefficients are >0.7 (Methods). further additions of traits to our data would not substantially alter We found that highly significant modularity is still present in each our results. We thus predict that the median number of traits af- of the five gene–trait networks (Table S3). fected by a gene is no greater than a few percent of the total number of traits in an organism. Furthermore, because gene Genes Affecting More Traits Have Larger Per-Trait Effects. The yeast pleiotropy is largely owing to the involvement of the same mo- morphological pleiotropy data contain quantitative information lecular function in multiple different biological processes rather about the phenotypic effect size of mutations, which is another than the presence of multiple molecular functions per gene (26), important parameter in genetics that has never been available at random mutations in a gene will likely affect the same traits as the the genomic scale.